I timed a bunch of Linux kernel builds on my Core 2 Quad Xeon box, and found that I got the fastest builds with -j 8. One would reasonably expect that you'd get the best time with the same number of jobs as cores, but my testing showed that was clearly not the case. I didn't look into it, but I expect that the reason is that parts of the build are CPU bound, while other parts of the build are I/O bound. A job will just block if it has to wait for the disk, and so will cease doing any useful work. If you then have a CPU bound job ready to run, useful work proceeds. I eagerly await the day when I have the cash to buy my second Core 2 Quad. My Supermicro X7DWA-N motherboard is socketed for two Xeons, but works fine with just one installed. So when I built the box, I spent as much money as I could on providing a future upgrade path. FB-DIMM memory though... that's gonna bankrupt me. Mike -- Michael David Crawford mdcrawford at gmail dot com GoingWare's Bag of Programming Tricks http://www.goingware.com/tips/